The Maze: Search is not moving from one universal box to another. It is splitting into destinations with different strengths and intents. Similarweb’s April-to-July index shows Claude and Amazon.com gaining while Google remains almost flat, ChatGPT softens, Gemini gives back an early rise, and Perplexity falls sharply. That is less a coup against Google than a portfolio shift: people can discover on Google, reason in an assistant, create with another tool, and buy on a marketplace without treating those actions as one journey.
Claude and Amazon are the only two clear gainers in the July reading. Claude rises from an April baseline of 100 to roughly 112 in May, 115 in June and 111 by July 19. Amazon.com moves from 100 to about 102, 107 and 104. The paths are different in size but similar in shape: both peak in June and retain part of the gain. That matters because one is an AI assistant and the other is a product marketplace. Attention is not simply migrating from “old search” to “AI search.” It is accumulating around tools that own a useful job.
The losers do not share one explanation. Google stays near 100 through June and ends around 99, which looks more like stability than displacement. ChatGPT sits near 98 in May and June before easing to roughly 93. Gemini climbs to about 104, then falls to 94. Perplexity drops every month, ending around 64. Those paths could reflect product releases, seasonality, distribution changes or measurement noise. What they do show is divergence: six discovery surfaces are moving on different rhythms rather than behaving as one substitute market.
The index is about momentum, not scale. Every platform starts at 100 against its own April traffic, so a rising smaller destination can still have far fewer visits than a flat larger one. July is also month-to-date through the 19th, while the prior months appear complete. The post’s “Claude +15%” and “Amazon +8%” claims fit the June peaks better than the July endpoints. That distinction prevents a common analytical mistake: converting a normalized growth line into a market-share story the source does not support.
Visibility now has to be measured by job and outcome. Similarweb’s broader 2026 report describes a fragmenting AI-search market in which referrals, citations and ad placements follow different rules. Its May analysis also found ChatGPT-referred visits converting at 7.1% in its April–May panel, behind paid search at 7.8% but ahead of the other listed channels. Traffic momentum tells operators where attention is moving; referral quality and conversion tell them whether that attention pays.
Why it matters: Ecommerce teams built the last playbook around one dominant doorway: rank on Google, buy search ads, and send shoppers to a product page. The new system looks more like a web. A customer may discover a need on Google, compare options in ChatGPT, build a plan in Claude and finish on Amazon. Brands therefore need platform-specific visibility, structured product information and separate referral measurement. The goal is no longer to win one ranking. It is to remain findable and credible at every handoff where intent changes shape.
